I Think I Love Deepseek R1

00:13:25
https://www.youtube.com/watch?v=GNHuJE_GARc

Zusammenfassung

TLDRThe speaker shares their excitement for R1, primarily because it allows users to run AI models locally, offering independence from companies like OpenAI which they distrust for data collection practices. They find the ability to set up personal AI systems using affordable hardware like Mac Minis and GPUs a significant improvement. Though they foresee rising prices in GPU technology as more users gain access, they believe this shift is empowering for individual users who want to integrate AI into their projects. The speaker is eager to explore the potential for offline AI models and the possibilities this technology allows for personal expression and control over data.

Mitbringsel

  • 🚀 Emphasis on local AI model usage instead of cloud dependency.
  • 🛡️ Importance of data privacy in AI applications.
  • 💻 Plans to build a personal AI setup with Mac Minis and GPUs.
  • 📈 Beliefs about increasing hardware prices due to greater accessibility.
  • 🎯 Excitement about integrating personal data into AI experiments.
  • 💡 Skepticism towards OpenAI's intentions and practices.
  • ✨ Vision for a future with decentralized control of AI.
  • 🔍 Interest in using offline models to avoid data tracking.
  • 🤖 Critique on AI's code generation abilities.
  • 🎉 General enthusiasm for possibilities that R1 introduces.

Zeitleiste

  • 00:00:00 - 00:05:00

    The speaker expresses excitement about the R1 model, emphasizing its accessibility for individuals to run models locally on less expensive hardware. They appreciate the ability to maintain control over their data without relying on corporations like OpenAI, which they accuse of mishandling personal information. The speaker plans to create a local setup for experimenting with AI models, underscoring the potential for independence from larger companies and the ability to innovate personally.

  • 00:05:00 - 00:13:25

    The discussion extends to the anticipated increase in GPU prices due to wider accessibility and interest in AI technology, suggesting that this is a positive development. The speaker remains skeptical about the intentions of companies like OpenAI, viewing them as self-serving. They envision a future where offline models can be used effectively, facilitating personalized interactions with AI and a more favorable development landscape beyond the realm of corporate control.

Mind Map

Video-Fragen und Antworten

  • What excites the speaker about R1?

    The potential to run smaller AI models locally without relying on companies like OpenAI.

  • Why does the speaker distrust OpenAI?

    They believe OpenAI collects user data and is not genuinely focused on the good of humankind.

  • What hardware does the speaker plan to use?

    The speaker plans to use Mac Minis and GPUs to build a personal AI setup.

  • What are the speaker's views on AI prices?

    They believe GPU prices will continue to rise as more individuals begin using AI.

  • What kind of projects does the speaker want to pursue with AI?

    They hope to integrate their own models and data to create custom solutions.

  • How does the speaker feel about AI coding capabilities?

    They think AI produces poor code and prefer traditional coding methods.

  • What is the speaker's sentiment towards the future of AI?

    They are very excited about the democratization of AI access and local use.

  • Does the speaker plan to use online AI services?

    No, they prefer offline models to maintain control over their data.

Weitere Video-Zusammenfassungen anzeigen

Erhalten Sie sofortigen Zugang zu kostenlosen YouTube-Videozusammenfassungen, die von AI unterstützt werden!
Untertitel
en
Automatisches Blättern:
  • 00:00:00
    I'm extremely excited about R1 not for
  • 00:00:02
    the reasons why other people are excited
  • 00:00:04
    for R1 Okay the reason why I'm excited
  • 00:00:06
    for R1 is actually quite simple when I
  • 00:00:09
    look at all this stuff what I see is I
  • 00:00:11
    see the availability for us to be able
  • 00:00:13
    to actually have a model that is really
  • 00:00:15
    really good that we can actually use
  • 00:00:17
    that you and I can actually use and put
  • 00:00:20
    to use even some of the smaller models
  • 00:00:22
    yeah they're not as good but we can
  • 00:00:24
    actually use it and for me that is
  • 00:00:26
    really really amazing that's what I
  • 00:00:28
    absolutely love like I can go set up
  • 00:00:31
    some sort of basic couple gpus and be
  • 00:00:33
    able to run even a decent sized model
  • 00:00:35
    it's not a great model right I don't
  • 00:00:37
    need to spend $ 25,000 you could spend
  • 00:00:38
    $1,000 get a good enough like piece of
  • 00:00:41
    Hardware to be able to run something
  • 00:00:42
    small and local it can be completely
  • 00:00:44
    offline and you can have it just for you
  • 00:00:46
    and be able to run it without having to
  • 00:00:48
    worry about all the other stuff the data
  • 00:00:49
    collection all that that actually seems
  • 00:00:52
    pretty amazing like that actually seems
  • 00:00:54
    like a huge W and so I'm pretty happy
  • 00:00:57
    about that uh and that's actually what I
  • 00:00:59
    want to more of is more of that more
  • 00:01:03
    ability to be able to run stuff myself
  • 00:01:06
    without having to r on open AI tracking
  • 00:01:09
    and logging me okay because they already
  • 00:01:11
    steal all of our information they
  • 00:01:13
    already do all of that that stuff and
  • 00:01:15
    it's actually pretty awful and so the
  • 00:01:17
    fact that I can take control I can be
  • 00:01:19
    the one like open AI is not open open AI
  • 00:01:21
    is not the one that has your back okay
  • 00:01:24
    they don't they don't they don't have
  • 00:01:26
    your back they they just want your data
  • 00:01:28
    and they want to take all the money for
  • 00:01:30
    from you okay whereas this one actually
  • 00:01:32
    feels pretty awesome now the whole
  • 00:01:33
    Chinese government and all that that's
  • 00:01:34
    why I would just never want to use it
  • 00:01:36
    online like I'm not going to use their
  • 00:01:37
    online interface I I just assume that
  • 00:01:39
    they're no better than open AI they're
  • 00:01:41
    using it all to take data and to be able
  • 00:01:43
    to try to attempt to continue on using
  • 00:01:46
    it as training and all that and who
  • 00:01:48
    knows what else they're actually using
  • 00:01:49
    the data for like I'm totally and
  • 00:01:51
    completely not buying any of it at all
  • 00:01:55
    for whatever reason so with that in mind
  • 00:01:58
    I do love that I am now the owner and so
  • 00:02:01
    this is why I'm buying a couple of those
  • 00:02:03
    Mac minis I'm going to try to build my
  • 00:02:04
    own Little Mac Mini Farm I'd like to
  • 00:02:06
    also play around with a couple gpus and
  • 00:02:08
    build like a motherboard with a few
  • 00:02:09
    different gpus right you can get a lot
  • 00:02:11
    of like pcie Express slots and then
  • 00:02:14
    pcie PCI Express that's kind of funny
  • 00:02:16
    PCI uh Express slots Jam a couple of
  • 00:02:19
    those in and see how it goes see how
  • 00:02:21
    we're producing be able to compare them
  • 00:02:23
    and be like this is what it cost to do
  • 00:02:24
    this level this is what it cost to be
  • 00:02:26
    able to do this level and then as the
  • 00:02:27
    models inevitably improve at least my
  • 00:02:30
    Hardware despite the fact that my
  • 00:02:31
    Hardware hasn't changed I might get some
  • 00:02:33
    improvement in the models which could be
  • 00:02:35
    amazing and then on top of that you
  • 00:02:37
    could refresh your own hardware and kind
  • 00:02:39
    of be able to have the level that you
  • 00:02:41
    want to be able to uh spend at which
  • 00:02:43
    does kind of feel pretty cool and then
  • 00:02:45
    even more so if you're at a company that
  • 00:02:46
    doesn't allow AI training and all that
  • 00:02:48
    you could actually have your own offline
  • 00:02:50
    one if you really wanted it and like I
  • 00:02:52
    mean that would be awesome right like
  • 00:02:54
    that' be pretty dang awesome I really
  • 00:02:56
    love this idea of being able to play
  • 00:02:58
    around to be able to play around with
  • 00:02:59
    your own model right you're not going to
  • 00:03:01
    run the 7 uh 700b model uh probably not
  • 00:03:04
    I probably won't be able to afford that
  • 00:03:05
    because that's going to take a lot of
  • 00:03:06
    money to be able to do that like a lot a
  • 00:03:08
    lot of money but I'm going to be able to
  • 00:03:10
    get something pretty close I'm going to
  • 00:03:11
    be able to get a pretty big model being
  • 00:03:12
    able to run locally and that's going to
  • 00:03:14
    be awesome and so for me what I see is
  • 00:03:17
    that I didn't actually think I was going
  • 00:03:19
    to see this with AI honestly I always
  • 00:03:21
    thought the future of AI was going to be
  • 00:03:23
    strictly more control by fewer companies
  • 00:03:27
    but it might not be that there might be
  • 00:03:29
    a small potential future in which it's
  • 00:03:34
    not strictly ran by a bunch of companies
  • 00:03:37
    or by a few companies and to me that
  • 00:03:39
    could be very very exciting say thank
  • 00:03:41
    you China uh that's kind of hard for me
  • 00:03:44
    to say thank you China thank you the
  • 00:03:46
    people from China uh so you know it's
  • 00:03:49
    just something to kind of think about I
  • 00:03:51
    really hope so um maybe I don't I I
  • 00:03:54
    don't know what they can and can't run I
  • 00:03:55
    still got to do the exploration myself
  • 00:03:57
    so I'm not going to be someone that's
  • 00:03:58
    going to tell you all the things yet
  • 00:03:59
    because I haven't I haven't been able to
  • 00:04:02
    do it yet so when I when I get out there
  • 00:04:05
    and I'm going to try to really work
  • 00:04:07
    these models and see how they actually
  • 00:04:08
    go I think it's going to be a lot of fun
  • 00:04:10
    because I think one thing that's going
  • 00:04:11
    to be super fun is being able to
  • 00:04:12
    integrate your editor with your own
  • 00:04:14
    models right like if I'm able to do any
  • 00:04:16
    form of training at all on these or any
  • 00:04:18
    any type of any type of rag or whatever
  • 00:04:21
    all the things are that people are use
  • 00:04:22
    or any of the new future techniques we
  • 00:04:24
    won't know about be able to provide my
  • 00:04:26
    own set of data for them to be able to
  • 00:04:27
    work off of you could imagine that you
  • 00:04:29
    could actually make a really amazing
  • 00:04:31
    experience and that could be pretty cool
  • 00:04:33
    I don't think GPU prices are going down
  • 00:04:35
    at all I think this the only thing that
  • 00:04:37
    I see is that we are already going to
  • 00:04:39
    have increased prices because more and
  • 00:04:41
    more individuals are now able to opt in
  • 00:04:44
    to using these things I don't see
  • 00:04:46
    somehow gpus and stuff going down I see
  • 00:04:49
    gpus and stuff going up because even
  • 00:04:51
    more people have access it's not just
  • 00:04:53
    stuck to some small tiny fraction of
  • 00:04:56
    people because right now all that's
  • 00:04:57
    happening is like Facebook Amazon Google
  • 00:05:01
    open AI they're just like come on please
  • 00:05:04
    please Nvidia give me all your gpus like
  • 00:05:06
    they're the ones that are just funneling
  • 00:05:08
    this current market next we'll be able
  • 00:05:10
    to have many people be able to do it so
  • 00:05:13
    very exciting I think GPU prices are
  • 00:05:15
    going to go just further up I I think
  • 00:05:17
    that's all we're going to see so I'm
  • 00:05:19
    very excited about this honestly I'm
  • 00:05:20
    very excited and and lastly I think this
  • 00:05:22
    is a really great because open AI always
  • 00:05:25
    claims to be wanting to do things for
  • 00:05:27
    the good of humankind and I don't think
  • 00:05:29
    open AI has any intent on doing things
  • 00:05:32
    good for the uh for humankind for if
  • 00:05:34
    they were trying to do stuff for the
  • 00:05:36
    good of humankind then I think they
  • 00:05:39
    would have done more stuff open source
  • 00:05:41
    and I do not believe at any point have
  • 00:05:45
    they done anything that's for the good
  • 00:05:46
    of humankind I think they're truly
  • 00:05:48
    trying to take advantage and make us all
  • 00:05:50
    dependent on their stuff and be able to
  • 00:05:52
    do that and so I am very happy that a
  • 00:05:56
    beautiful beautiful piece of Open Source
  • 00:05:58
    came out very happy about it and there
  • 00:06:00
    you go that's like my true take they're
  • 00:06:02
    going to rewrite the social contract for
  • 00:06:04
    the better like I mean like the things
  • 00:06:06
    the way that Sam Alman talks and all
  • 00:06:08
    that should be kind of alarming to a lot
  • 00:06:10
    of people right like he really he wants
  • 00:06:13
    to rewrite the social contract and I'm
  • 00:06:15
    going I'm going to give you a little
  • 00:06:16
    hint about anything that has to do when
  • 00:06:19
    somebody of great power says this they
  • 00:06:21
    want to rewrite whatever World such that
  • 00:06:24
    they get more power and you get less
  • 00:06:26
    that's what that means this is true in
  • 00:06:29
    all types of government for all time
  • 00:06:31
    anybody with power that can rewrite
  • 00:06:34
    contracts will rewrite it in their favor
  • 00:06:37
    this is not some uniquely American
  • 00:06:39
    aspect just read a little bit of history
  • 00:06:42
    and it's every single kingdom is always
  • 00:06:44
    pegging towards one person gaining all
  • 00:06:46
    the power and trying to kill everybody
  • 00:06:48
    it's just nuts and so I really have
  • 00:06:50
    always hated open AI it's actually part
  • 00:06:51
    of the reason why I've been so obstinate
  • 00:06:53
    with AI is because I've always found
  • 00:06:55
    them to just give me just like
  • 00:06:57
    theof you know what I mean lastly I see
  • 00:07:00
    a bunch of people saying a bunch of
  • 00:07:01
    stupid stuff in chat okay hey sorry that
  • 00:07:02
    a bunch of people are saying stupid
  • 00:07:04
    stuff in uh chat okay you know what
  • 00:07:07
    pegging means I don't mean pegging in
  • 00:07:08
    that kind of sense shut up uh but lastly
  • 00:07:11
    yeah like I don't trust deep seek that's
  • 00:07:13
    why I want to run run it offline I don't
  • 00:07:15
    want to give it access to the internet
  • 00:07:17
    because I don't know if it calls home or
  • 00:07:18
    not I don't know what it does I'm
  • 00:07:21
    probably too stupid to actually be able
  • 00:07:23
    to go over all of that so I'm just not
  • 00:07:25
    going to go over it instead if I can
  • 00:07:27
    have a completely offline model I will
  • 00:07:29
    have my own offline model doing my own
  • 00:07:31
    offline things by myself that will be
  • 00:07:33
    fantastic that's like exactly what you
  • 00:07:35
    want to see in the future I don't want
  • 00:07:37
    that China phone home okay we don't want
  • 00:07:39
    that it calls home for sure it very well
  • 00:07:42
    could like I don't know again this is
  • 00:07:45
    why I want offline model I found the USB
  • 00:07:48
    on the street should I plug it into my
  • 00:07:49
    computer well does it say family photos
  • 00:07:51
    because then you could look at who's in
  • 00:07:52
    the photos and find out who has who lost
  • 00:07:54
    their USB anyways recalling home well
  • 00:07:57
    did someone actually find that it calls
  • 00:07:58
    home uh user when you uh use our
  • 00:08:00
    services we may collect your text and
  • 00:08:02
    audio input prompt uploaded files
  • 00:08:04
    feedback chat history or other content
  • 00:08:05
    you provide to our model and service
  • 00:08:08
    yeah I'm curious what that means to our
  • 00:08:09
    model and service is it model and
  • 00:08:11
    Service as one Unity one unit because if
  • 00:08:13
    that's the case that makes that make
  • 00:08:15
    that should be not surprising to anybody
  • 00:08:16
    if you are sending data to a website
  • 00:08:19
    they're collecting that data in some
  • 00:08:21
    level if you are using a model are they
  • 00:08:24
    collecting the data do they really have
  • 00:08:26
    the power to send home how is it
  • 00:08:27
    happening right those are very different
  • 00:08:29
    those are different ones so when it says
  • 00:08:30
    model and service does it mean the
  • 00:08:33
    previous stuff applies to the model and
  • 00:08:36
    or to the service is that an or or is
  • 00:08:39
    that an and because when you when you
  • 00:08:40
    program you use ores and when you speak
  • 00:08:42
    in English you mean and and often it can
  • 00:08:45
    be kind of confusing to to work between
  • 00:08:47
    the two when you're when you're
  • 00:08:49
    translating from logic to English well
  • 00:08:52
    what I mean by that is like let me give
  • 00:08:54
    you an example let's make a function of
  • 00:08:56
    food that I like if I say the phrase I I
  • 00:08:59
    like pizza and burgers you would write
  • 00:09:01
    that function as return pizza or Burger
  • 00:09:06
    you wouldn't say pizza and burger
  • 00:09:07
    because a food can't be you know okay
  • 00:09:10
    maybe they can be both right but
  • 00:09:13
    typically you know what I'm trying to
  • 00:09:14
    say there's this thing that happens in
  • 00:09:16
    the translation between what they're
  • 00:09:18
    saying versus what they mean and I don't
  • 00:09:21
    know what they mean when I see and does
  • 00:09:24
    it mean provided to our
  • 00:09:27
    model and service hasn't it's either the
  • 00:09:30
    model or the service and they both go up
  • 00:09:32
    or does it mean it can only be through
  • 00:09:35
    the service to the
  • 00:09:36
    model right I don't know if it's a
  • 00:09:39
    contingent and do you see what I'm
  • 00:09:41
    saying right
  • 00:09:43
    anyways now that we got that now that we
  • 00:09:45
    got that out of the way but I do want to
  • 00:09:47
    I do want to say this is that I I do
  • 00:09:48
    think that this could actually spell a
  • 00:09:50
    huge cool new world and what I mean by
  • 00:09:53
    that is that like right now if you ask
  • 00:09:56
    Chad jiy to do Zig stuff it is just God
  • 00:09:58
    awful okay it suggests it suggests
  • 00:10:01
    built-ins that don't exist it constantly
  • 00:10:04
    does stuff that can't even be real it
  • 00:10:06
    just it's just constantly wrong it is
  • 00:10:08
    very very bad at even simple examples
  • 00:10:11
    whereas if I can take my Ai and I can
  • 00:10:14
    have it on what I want it to be more
  • 00:10:16
    focused on I can start getting it to
  • 00:10:19
    answer my questions based on my
  • 00:10:21
    documents that I'm providing which means
  • 00:10:23
    that I can literally take these
  • 00:10:24
    documents and I can
  • 00:10:26
    say based on these documents please tell
  • 00:10:28
    please show me these spots in which I
  • 00:10:30
    can read about how to do this specific
  • 00:10:32
    action and then can you give me an
  • 00:10:33
    example based off these right like you
  • 00:10:35
    could do some pretty cool stuff you can
  • 00:10:37
    do that with chat jiy too you can do
  • 00:10:38
    that with chat jiy too but again I don't
  • 00:10:40
    want to just give all my data to open AI
  • 00:10:42
    to tell you the truth I don't like open
  • 00:10:43
    AI I don't trust them so docs like and
  • 00:10:45
    cursor cursor does a good job with this
  • 00:10:46
    so if you could use cursor plus your own
  • 00:10:48
    offline model would that not be awesome
  • 00:10:52
    right if I can ask cursor questions in
  • 00:10:54
    my offline model and be able to have
  • 00:10:55
    examples produced for me that could
  • 00:10:57
    actually that could actually make me
  • 00:10:58
    really like The Experience anyways just
  • 00:11:01
    a thought that's a startup coming soon I
  • 00:11:02
    know I want to I want to do something
  • 00:11:04
    along those lines because I think it's
  • 00:11:05
    very exciting anyways I'm very excited
  • 00:11:06
    for the future of AI I think this is
  • 00:11:08
    actually the most exciting this is the
  • 00:11:10
    most exciting development of R1 or of AI
  • 00:11:14
    in my personal opinion I thought Chad
  • 00:11:16
    jip I thought co-pilot was probably the
  • 00:11:18
    first most exciting thing ever I still
  • 00:11:20
    remember the magic I felt when co-pilot
  • 00:11:23
    finished my if else and did the else
  • 00:11:26
    correctly with a tie which did like if
  • 00:11:28
    player one wins and then it kind of
  • 00:11:30
    helped me with player two and then just
  • 00:11:31
    fully eled everything and matched kind
  • 00:11:34
    of the if if else above it I still
  • 00:11:36
    remember looking at that going this was
  • 00:11:38
    one of the most amazing experiences I've
  • 00:11:40
    ever had absolutely love this I am so
  • 00:11:44
    like stoked about that and so when I see
  • 00:11:46
    this I actually think that this is in
  • 00:11:48
    fact a more magic thing that has
  • 00:11:51
    happened because this gives me the tools
  • 00:11:53
    to create that and I'm so excited about
  • 00:11:55
    it so there you go that's how I
  • 00:11:58
    feel I'm very excited now I've actually
  • 00:12:01
    now I'm becoming an AI Fanboy have I
  • 00:12:03
    just become an AI Tech bro did I just
  • 00:12:05
    become an AI Tech bro maybe maybe maybe
  • 00:12:10
    maybe no probably not but you could
  • 00:12:12
    imagine that at least I'm pretty excited
  • 00:12:14
    you know okay I'm not a stop it I'm not
  • 00:12:17
    goingon to be a I'm not gonna be a AI
  • 00:12:19
    Tech bro uh I still like writing code
  • 00:12:21
    that's it I still think that AI produces
  • 00:12:23
    shitty code okay I can't help it it prod
  • 00:12:26
    it produces just the worst code in the
  • 00:12:28
    universe I I don't know how you guys
  • 00:12:30
    keep do it I I honestly don't understand
  • 00:12:32
    why you guys use it all the time to me
  • 00:12:33
    it's shocking it's like it's it's
  • 00:12:35
    honestly personally shocking but hey
  • 00:12:36
    that's you guys you guys can use it all
  • 00:12:39
    you want that's all I got to say I can
  • 00:12:41
    produce worst code I actually am not
  • 00:12:43
    sure if that's true it's okay at
  • 00:12:45
    debugging you know what's you know who's
  • 00:12:47
    much better at debugging you when you
  • 00:12:49
    get good at debugging you're not like
  • 00:12:51
    just like a little bit better you're
  • 00:12:53
    like massively better it's like crazy
  • 00:12:55
    how much better you can actually be by
  • 00:12:57
    getting good anyways you know who's
  • 00:12:59
    better your mom damn you guys got me on
  • 00:13:01
    that one you guys got me anyways the
  • 00:13:08
    name I hate you guys what am I saying
  • 00:13:11
    you know what you don't even you don't
  • 00:13:12
    you know what why don't you press the
  • 00:13:14
    like button and the Subscribe button
  • 00:13:16
    okay how about you show how about you
  • 00:13:17
    give me something this time why don't
  • 00:13:19
    you give me a little why you give me a
  • 00:13:20
    little goodness okay okay just put give
  • 00:13:23
    it to me a Jen
Tags
  • R1
  • AI models
  • local use
  • OpenAI
  • data privacy
  • hardware
  • democratization
  • technology
  • offline models
  • personal projects